Texture-preserving abstraction

Image abstraction traditionally eliminates texture, but doing so ig-nores the more elegant alternative of texture indication, e.g., sug-gesting the presence of texture through irregular silhouettes and lo-cally chosen details. We propose a variant of geodesic image filter-ing which preserves the locally strongest edges, leading to preser-vation of both strong edges and weak edges depending on the sur-rounding context. Our contribution is to introduce cumulative range geodesic filter-ing, where the distance in the image plane is lengthened propor-tional to the color distance. We apply the new filtering scheme to abstraction applications and demonstrate that it has powerful structure-preserving capabilities, especially regarding preservation and indication of textures